A statistical model is proposed to study the fracture of graphite. Thi
s model, based on a more general one proposed by She et al., uses a lo
cal fracture criterion for a microcrack, a distribution function for m
icrocracks and the weakest link principle to predict failure probabili
ty as a function of applied loading and distribution of microcracks. T
he model considers the effect of the three-dimensional stress state an
d therefore gives a general representation of both the effects of comp
lex loading and microcrack distribution. The inputs to the model can b
e determined by either studying the microstructural features of the gr
aphite or by choosing inputs to make the model prediction fit a set of
experimental results. The latter method is used to apply the model to
failure results of Rose and Tucker. One set of results is used to cal
ibrate the model which is then applied to predict the failure behavior
of a second set of experimental results. Finally, the effect of the v
arious input variables on failure probability is studied first by cons
idering a graphical representative of failure probability as a functio
n of input variables and second by writing the equations in terms of n
ondimensional variables.